Date: Mon, 30 Jan 2012 02:45:04 -0500
Reply-To: Søren Lassen <s.lassen@POST.TELE.DK>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Søren Lassen <s.lassen@POST.TELE.DK>
Subject: Re: non-SAS-killing ABORT for interactive mode, is there one?
Content-Type: text/plain; charset=ISO-8859-1
Yevgeniy,
I made a small mistake in my orginal post. The macro variable that I
had meant to use was SYSCC, not SYSRC. SYSCC works the way I described,
while SYSRC reports errors with the X statement.
Sorry about that.
Regards,
Søren
On Thu, 26 Jan 2012 10:19:57 -0500, Bolotin Yevgeniy
<YBolotin@SCHOOLS.NYC.GOV> wrote:
>This is pretty cool, will have to play with it
>
>Thanks!
>
>
>
>-----Original Message-----
>From: Søren Lassen [mailto:s.lassen@POST.TELE.DK]
>Sent: Thursday, January 26, 2012 5:53 AM
>To: SAS-L@LISTSERV.UGA.EDU; Bolotin Yevgeniy
>Subject: Re: non-SAS-killing ABORT for interactive mode, is there one?
>
>Yevgeniy,
>Here is a solution that uses old-style macros (mixed with "new"-style):
>%macro errchk;
> %if &sysrc %then %do;
> macro run run cancel%
> macro sql sql noexec%
> macro kill%
> %end;
> %else %do;
> macro run run%
> macro sql sql%
> macro kill kill%
> %end;
>%mend;
>
>%let sysrc=0; /* no errors yet */
>%errchk;
>
>data _z; a = 1; output; a = 2; output; run;
>
>%let sysrc=1; /* Simulate an error */
>%errchk;
>
>Proc sql;
> create table _z as select * from _z, _z;
>quit;
>
>proc datasets lib=work kill nolist mt=data;
>run;
>
>
>What it does: it checks the automatic variable SYSRC. SYSRC is set to the
>maximum of the automatic variable RC (I think it is that one) and SYSRC
>after each step. However, you can reset it to 0 in your program.
>
>If SYSRC is not 0, it makes SAS do an old-style macro substitution of
>the word RUN with RUN CANCEL, of SQL with SQL NOEXEC and of KILL with
>nothing. If SYSRC is 0, it resets the tokens to their normal values.
>
>This means that after the (simulated) error, the SQL procedure statements
>are not executed, only syntax checked. And likewise are datasteps
>and most other procedures that excute after RUN; an exception is PROC
>DATASETS, which deletes tables (and/or views,catalogs) regardless,
>when the KILL options is used - therefore the KILL token is nulled when
>there is an error.
>
>Just remember to use RUN in PROC DATASETS and not QUIT. And generally,
>use RUN after all your datasteps/procedures except SQL.
>
>If you need to stop execution based on data values, you can do something
>like:
>data _null_;
> set _z;
> if a>1 then do;
> call symput('sysrc','1');
> stop;
> end;
>run;
>
>%errorchk;
>
>Regards,
>Søren
>
>On Wed, 25 Jan 2012 15:58:09 -0500, Bolotin Yevgeniy
><YBolotin@SCHOOLS.NYC.GOV> wrote:
>
>>Joe and _null_:
>>
>>Abort abend/return does exactly what I want, but with one exception: it
>>ALSO kills SAS (as in, the application closes). I want to retain my SAS
>>session (and formats, macro variables/sasmacr, WORK library, etc)
>>
>>option OBS = 0 kinda works, but not the way I need (it goes through the
>>rest of the code anyway, nuking *everything* in its path)
>>
>>
>>
>>so here's a goal: given the following (ultra-simplified) code
>>
>> data _z; a = 1; output; a = 2; output; run;
>> * CODE TO STOP EXECUTION HERE;
>> Proc sql; create table _z as select * from _z, _z; quit;
>>
>>I want to run the whole thing, and end up with the dataset _z with two
>>observations (and not zero)
>>
>>This is obviously a silly example, but I may end up using this break
>>logic in the middle of a large set of macro calls, where datasets are
>>updated left and right (some of these files are 15gb - so during
>>different stages of processing I may merge it to other files maybe a
>>dozen times each - making a differently-named version each time is not
>>feasible)
>>
>>
>>
>>
>>
>>
>>
>>-----Original Message-----
>>From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU] On Behalf Of
>>Data _null_;
>>Sent: Wednesday, January 25, 2012 2:40 PM
>>To: SAS-L@LISTSERV.UGA.EDU
>>Subject: Re: non-SAS-killing ABORT for interactive mode, is there one?
>>
>>Seems like in DMS that ERRORABEND may work toward this objective too.
>>
>>On 1/25/12, Joe Matise <snoopy369@gmail.com> wrote:
>>> You need to set obs=0 also. This is what batch mode does when it
>>encounters
>>> an error - sets SYNTAXCHECK and obs=0.
>>>
>>> -Joe
>>>
>>> On Wed, Jan 25, 2012 at 1:04 PM, Bolotin Yevgeniy
>>> <YBolotin@schools.nyc.gov>wrote:
>>>
>>>> Syntaxcheck doesn't seem to be doing what I want it to be doing
>>>> (DMSSYNCHK is the interactive version) - it only affects the dataset
>>>> with the error in it
>>>>
>>>> Both valid datasteps and proc sql execute just fine in the below
>>sample
>>>>
>>>>
>>>>
>>>> option DMSSYNCHK SYNTAXCHECK;
>>>>
>>>> data _null_;
>>>> set non_existent_dataset;
>>>> run;
>>>>
>>>> data _null_;
>>>> kjhlksafhkdfhlah; /* syntax error */
>>>> run;
>>>>
>>>> data _z;
>>>> a = 1;
>>>> run;
>>>>
>>>> data _z2;
>>>> set _z;
>>>> run;
>>>>
>>>> proc sql;
>>>> create table _z3 as
>>>> select * from _z;
>>>> quit;
>>>>
>>>>
>>>>
>>>>
>>>> -----Original Message-----
>>>> From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU] On Behalf Of
>>>> Data _null_;
>>>> Sent: Wednesday, January 25, 2012 1:16 PM
>>>> To: SAS-L@LISTSERV.UGA.EDU
>>>> Subject: Re: non-SAS-killing ABORT for interactive mode, is there
>>one?
>>>>
>>>> You could put SAS into syntax check, obs=0 mode based on some
>>>> condition. You will need to check the exact options that need to be
>>>> set.
>>>>
>>>> On 1/25/12, Bolotin Yevgeniy <YBolotin@schools.nyc.gov> wrote:
>>>> > This has probably been asked before but I can't seem to find what I
>>>> > want...
>>>> >
>>>> > Is there a simple command I can write in code to force SAS to abort
>>>> the
>>>> > currently submitted code (i.e. equivalent to Cancel Submitted
>>>> Statements
>>>> > from the Break menu), that DOESN'T actually shut down SAS in the
>>>> process
>>>> > like ABORT does?
>>>> > Even if the abort cannot be made conditional, this can work wonders
>>>> for
>>>> > debugging (i.e. insert as a breakpoint)
>>>> >
>>>> >
>>>> > I've seem some interesting approaches to doing this with macros
>>>> > injecting comment delimiters (i.e. /* ) but this is both way too
>>>> > complicated for what I'm doing and requires way too much work to
>>>> > implement on existing projects
>>>> >
>>>> >
>>>> >
>>>> > Using SAS 9.2 on WinXP (32 bit)
>>>> >
>>>>
>>>
| 
